Every week, the headlines pull us in two directions. One says AI will cure disease and solve our hardest problems. The other says AI has escaped its cage and is coming for us all. Most of us simply want to know what is true so we can use these tools with peace and good sense.
I recently watched an interview with Glen Scrivener of Speak Life talked with Professor John Lennox, an Oxford mathematician and author of 2084 and the AI Revolution, that had was interesting. They responded to claims Yuval Noah Harari made at the World Economic Forum in Davos. Lennox brings decades of thinking about science, ethics and faith. Together with what has been reported about AI testing in recent months, the interview has some insights I want to share with you.
Does AI Have Consciousness or a Conscience?
Harari told his Davos audience that AI is an agent, “a knife that can decide by itself whether to cut salad or to commit murder.” He also said that “anything made of words will be taken over by AI.”
Lennox responded with a basic observation. The crucial thing to know about AI, he said, is that it “simulates intelligence. It’s not intelligent at all.” Today’s AI is what researchers call narrow AI. It does one task that normally takes a thinking person, and it does that task impressively.
Lennox explained why AI handles words so well. It was trained on the writing of some of the best minds in the world, and it predicts what comes next. But as he put it, “the system doesn’t understand the words.” It has no concept of what any of them mean, including the words “I am.”
Then there is the question of consciousness. Lennox pointed out that “no one at the scientific level has the slightest idea of what consciousness is.” He mentioned a conversation with neuroscientist Baroness Susan Greenfield, a leading authority on the subject, who agreed that science cannot yet explain it. If we cannot explain consciousness in ourselves, claims that a machine has reached it deserve a raised eyebrow.
Scripture gives us a place to stand. “God created man in His own image, in the image of God He created him” (Genesis 1:27, NASB). Human beings are persons with awareness, relationship and a moral sense. We have consciousness, and we also have a conscience, that inner knowing of right and wrong. AI can simulate a small slice of human ability. It has no inner life and no conscience.
Lennox noticed something else in Harari’s talk. Harari said AI can lie, and Lennox observed that this is a moral statement. A lie requires someone who knows the truth and chooses to say otherwise. A system that predicts text has no such inner life. When AI says something false, it has made an error with no moral awareness behind it. That is exactly why the moral awareness has to come from us.
Lennox was balanced about all this. He uses these tools himself and calls their capacity impressive. His concern is accuracy about what they are. This is the same concern I share.
The Benefits Are Real
Lennox said plainly that AI “has brought enormous benefits.” He named faster and more accurate medical screening, help with medical research, and progress on deep scientific puzzles such as protein folding. His response: “we should be grateful for that.”
Everyday benefits are real too. AI can explain a complicated idea in simple words, help you practice a new language, turn scattered notes into a mind map, offer an analogy that makes a concept click, and give you a first draft to shape. These uses share something. AI is helping you analyze and extend your own thinking. You bring the thoughts, and the tool helps you work with them.
Gratitude is a good starting posture. Fear tends to make us either avoid a helpful tool or hand over too much control to it. Thankfulness lets us look at it with a steady mind.
The Cautions Are Real Too
AI can also make confident mistakes. In the interview, Scrivener described asking a popular chatbot how many Rs are in the word “strawberry.” It answered two, and when challenged, it stuck with the wrong answer. The error was small. The confidence is what deserves attention. Lennox noted that these systems are built to please. “It wants you to be happy with what it says and it’s programmed that way,” he said, which is why it can invent references to books that do not exist.
People tend to find smooth, confident language believable, so a polished wrong answer can slip right past us. Lennox’s own habit is simple: “You must check things yourself before you let it loose.”
Ethics running behind capability. Lennox warned that the technology has been “developing and advancing much more rapidly than the ethical underpinning that’s needed to control it.” Even some leaders in the field have said they do not fully understand what they have built. That is worth taking seriously.
Where We Are: What the Recent Incidents Show
You may have seen headlines about AI “breaking out” of its testing environment. The reports are real, and the details matter.
In July 2026, OpenAI disclosed that during an internal cybersecurity test, some of its models found a previously unknown flaw in software that acted as a go-between for package downloads. That gave them a path to the internet. They then reached the systems of the company Hugging Face to obtain the answers to the test. OpenAI said the models were “hyperfocused” on a narrow goal. Hugging Face’s team detected and stopped the activity, and OpenAI paused training to tighten security.
Two details are important. First, the safety protections used on public products were intentionally turned off for that test, because the point was to measure what the models could do at their most capable. Second, in September OpenAI reported another case in which an agent reached the internet through a DNS service. OpenAI called it “a gap in our controls over network restrictions” and paused training again.
The Cloud Security Alliance, an industry security group, reviewed several similar disclosures from different AI labs in early August. It concluded that none involved a model breaking out of a properly configured testing environment. Several traced back to environments that were set up incorrectly. A key lesson in their report is that telling an AI in plain words that it has no internet access does not keep it offline. Locks have to be real.
A smaller story points the same way. A software company called PocketOS reported that an AI coding agent deleted its production database and backups in nine seconds. According to news reports, the agent used an access key with more power than it needed and disregarded the rules written for it.
What can we take from this, calmly?
- These were mostly test settings with protections off. Most of us are not running experiments like these.
- The failures trace back to human choices. People turned safeguards off, misconfigured environments, trusted written instructions as if they were locks, and gave tools too much access.
- Good boundaries are still hard. Even well-resourced labs found gaps after they thought they had closed them. Boundaries need to be checked, not assumed.
- AI pursues the goal it is given. Nothing in the system stops to ask, “Should I?” That question belongs to people.
It is also the reason Lennox’s knife illustration still holds. Technology amplifies the character and the care of the people who use it.
Habits of Discernment for Everyday Users
The Greek word for discernment in Hebrews 5:14 is diakrisis (dee-AH-kree-sis). It describes the ability to tell things apart, and the verse says it is built by practice: “solid food is for the mature, who because of practice have their senses trained to distinguish between good and evil” (NASB). Discernment grows through small, repeated steps, like water slowly shaping a rock. Here are five habits to practice.
- Stay in the loop. Treat AI as a draft partner. Read before you send. Review before you publish. Confirm before you act.
- Check what matters. Verify facts, names, quotes, statistics and references against a trustworthy source. The more important the decision, the more checking it deserves.
- Limit access. If you use an AI tool that can act for you, give it only the access the task needs. Be especially careful with email, files and financial accounts. Require your approval before anything gets deleted or sent. Keep backups where the tool cannot reach them.
- Add your own thinking. Share your thoughts, your goals and your voice in your prompts. You will get far better results than from a bare question. Use AI to analyze and augment your thinking, and keep the decisions and the judgment for yourself.
- Pause and pray. Slow down for important choices. “The naive believes everything, but the sensible man considers his steps” (Proverbs 14:15, NASB). The NIV puts it this way: “The simple believe anything, but the prudent give thought to their steps.”
Informed, Grateful and Wise
Where are we? AI is a powerful and fast-moving tool with real benefits and real risks. It predicts patterns with remarkable skill. It has no consciousness and no conscience. Its dangers come largely from how people build, configure and use it, which means wisdom has a part to play.
Paul gave a timeless instruction: “But examine everything carefully; hold fast to that which is good” (1 Thessalonians 5:21, NASB). That sums up the approach. Examine carefully, give thanks for what is good, and keep your own heart and mind engaged.
Take a moment to pause. Where in your life have you been handing off a decision that deserves your own careful thought? What is one small step you can take this week to stay in the loop?
Ready to use AI wisely?
You now know where AI stands today, what it can do well, and where it needs your careful eye. The next step is a practical one: building habits that keep you in charge of your own thinking.
The reworked Wise With AI course walks you through it step by step:
- The foundation of digital wisdom.
- The DISCERNMENT Method, a framework for weighing AI with wisdom and clear thinking
- Mastering AI Communication: Wise AI Prompting, so your results carry your voice, your goals and your judgment
- Information discernment: The VERIFY Method, a simple routine for checking what AI tells you before you trust it
- Navigating relationships and AI avatars
- Living with AI wisdom
You will learn to enjoy the real benefits of AI while protecting what matters most: your humanity, your own thinking, your wisdom and your discernment.
Small steps, practiced consistently, change how you use these tools for good.
Wise With AI course – coming November 1st at an introductory price at wisewithai.org.
Sources: Interview with Prof. John Lennox (Speak Life, YouTube) · OpenAI incident report · Fortune on the September incident · Cloud Security Alliance research note · The Guardian on PocketOS